Deyron Labs graphic: NVIDIA and Microsoft at Ignite, 19 Nov 2026; Vera Rubin NVL72 running in Azure; 10,240 GB200 GPUs in MLPerf Training v6.1; EY: 55% accuracy gain and 15x lower cost with Nemotron; Windows 365 Direct Stream preview

What happened

On 19 November 2026 NVIDIA published a blog post about its announcements with Microsoft at Microsoft Ignite. The post covers data-center hardware in Azure, MLPerf results, NVIDIA Nemotron models in Microsoft Foundry, and a Windows 365 feature. We read the post on 20 November 2026; the figures below are NVIDIA’s, and we have not seen Microsoft’s own announcement.

Key details

  • Hardware: NVIDIA says Vera Rubin NVL72 systems are “running in Azure data centers”.
  • MLPerf Training v6.1: Azure scaled to 10,240 NVIDIA GPUs using GB200 NVL72 systems on the DeepSeek-V3 benchmark, which NVIDIA describes as the largest scale in the round.
  • MLPerf Inference v6.1: Azure’s GB300 NVL72 systems led cloud submissions on DeepSeek-R1 interactive throughput by over 10%, according to Signal65 as cited by NVIDIA.
  • Nemotron in Foundry: Nemotron 3.5 Super and Lightning are available on Foundry Managed Compute; Nemotron 3.5 Bolt and Transcribe are listed as “coming soon”. NVIDIA says EY, using Programmed Training in Microsoft Foundry on Nemotron models, raised accuracy by 55% on its evaluation dataset and lowered cost 15x versus proprietary frontier models. That result is reported by NVIDIA and not independently verified.
  • Windows 365 Direct Stream: up to two 4K streams at 60 frames per second; a limited preview is open for sign-ups, with public preview planned for early 2027.
  • Not in the post: pricing, regional availability for Vera Rubin in Azure, and the terms of the EY comparison.

Why it matters

The post is a status update more than a launch: it says the next NVIDIA rack generation is already live in a major cloud and shows how Azure’s GB200 and GB300 systems perform in industry benchmarks. For customers, what matters next is when Vera Rubin capacity becomes rentable and at what price, neither of which the post states.

Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article.

Sources

  1. NVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCs Primary source , NVIDIA blog, 19 November 2026